IPM adoption, cooperative membership and farm economic performance: Insight from apple farmers in China

Wanglin Ma (Department of Global Value Chains and Trade, Lincoln University, Christchurch, New Zealand)
Awudu Abdulai (Department of Food Economics and Consumption Studies, University of Kiel, Kiel, Germany)

China Agricultural Economic Review

ISSN: 1756-137X

Article publication date: 6 September 2018

Issue publication date: 3 June 2019

Abstract

Purpose

The purpose of this paper is to examine the impact of agricultural cooperative membership on farmers’ decisions to adopt integrated pest management (IPM) technology and to estimate the impact of IPM adoption on farm economic performance.

Design/methodology/approach

An endogenous switching probit model that addresses the sample selection bias issue arising from both observed and unobserved factors is used to estimate the survey data from a sample of 481 apple households in China. A treatment effects model is employed to estimate the impact of IPM adoption on apple yields, net returns and agricultural income. In order to address the potential endogeneity associated with off-farm work variable in estimating both cooperative membership choice specification and IPM adoption specifications, a control function approach is used.

Findings

The empirical results show that cooperative membership exerts a positive and significant impact on the adoption of IPM technology. In particular, farmers’ IPM adoption decision is significantly associated with household and farm-level characteristics (e.g. education, farm size and price knowledge). IPM adoption has a positive and statistically significant impact on apple yields, net returns and agricultural income.

Practical implications

The findings indicate that agricultural cooperatives can be a transmission route in the efforts to proliferate the adoption and diffusion of IPM technology, and increased IPM adoption tends to improve the economic performance of farm households.

Originality/value

Despite the widespread evidence of health and environmental benefits associated with IPM technology, the adoption rate of this technology remains significantly low. This paper provides a first attempt by investigating to what extent and how agricultural cooperative membership affects IPM adoption and how IPM adoption influences farm economic performance.

Keywords

Citation

Ma, W. and Abdulai, A. (2019), "IPM adoption, cooperative membership and farm economic performance: Insight from apple farmers in China", China Agricultural Economic Review, Vol. 11 No. 2, pp. 218-236. https://doi.org/10.1108/CAER-12-2017-0251

Publisher

:

Emerald Publishing Limited

Copyright © 2018, Emerald Publishing Limited


1. Introduction

While pesticide use has increased agricultural production and productivity, its use, overuse and misuse have caused negative externalities on human health and the environment, as well as food safety (Dasgupta et al., 2007; Fernandez-Cornejo et al., 1998; Kabir and Rainis, 2014; Wilson and Tisdell, 2001; Xie et al., 2017). In particular, the overuse of chemical pesticides has led to pest resistance, resurgence and secondary outbreaks, which push farmers to use more new pesticides. To reduce the adverse effects of pesticide use, integrated pest management (IPM) technology has been introduced and implemented in agricultural production in many developing countries.

IPM refers to an ecologically based approach that makes the best use of all available technologies to manage pest problems sustainably. The primary objective of IPM technology is to minimize chemical pesticide use in relation to pest management, while maintaining or enhancing farm net returns with minimal environmental degradation. Previous studies have shown that IPM adoption significantly lowers pesticide use, saves production costs and maintains farm productivity for adopters (Carrión Yaguana et al., 2016; Cuyno et al., 2001; Dasgupta et al., 2007; Fernandez-Cornejo, 1996). In view of the significant benefits associated with IPM technology, many government and FAO programs have been developed to spread the technology. One such effective approach is the introduction of Farmer Field Schools (FFS) (Chhay et al., 2017; Kabir and Rainis, 2014; Sanglestsawai et al., 2015; Van Den Berg and Jiggins, 2007). However, IPM adoption rate remains low worldwide (Dasgupta et al., 2007; Kabir and Rainis, 2014). On the one hand, FFS is still not available in most regions (Kelly, 2005). On the other hand, due to low education levels, most small-scale farmers cannot understand the complex interrelationship between the pests/diseases existing in the cultivated crop and the knowledge-intensive IPM technology (Carrión Yaguana et al., 2016). Therefore, from a development policy perspective, it is particularly important to facilitate IPM adoption not only by FFS, but also through other institutional mechanisms.

Among agricultural programs, agricultural cooperative, as an important institutional innovation that promotes the adoption of agricultural technologies among smallholder farmers, has been well developed in developing countries (e.g. Abebaw and Haile, 2013; Ma et al., 2018; Verhofstadt and Maertens, 2014; Wossen et al., 2017). The studies by Abebaw and Haile (2013) for Ethiopia and Verhofstadt and Maertens (2014) for Rwanda have reported that cooperative membership exerts a positive and significant impact on adoption of pesticides with respect to pest management. Moreover, the existing literature has also recorded that agricultural cooperatives improve food safety and quality among members due to technical assistance (Moustier et al., 2010; Naziri et al., 2014). In their investigation of 60 farmer organizations in Vietnam, Naziri et al. (2014) found that farmer organizations provide members with technical assistance and monitoring for pest management, which help improve members’ food safety performance. Nevertheless, agricultural cooperative may play a much larger role in managing pest problems, since its goal in influencing agricultural production differs across countries and regions due to differences in natural resources and economic development conditions.

Given the importance of IPM adoption in minimizing pesticide use and the significant role of agricultural cooperatives in disseminating agricultural technologies and enhancing food safety practices, it is significant to understand whether cooperative organizations can promote IPM adoption. IPM is an information-intensive technology, and agricultural cooperatives may directly provide information to farmers through collective actions. However, there is lack of knowledge on how agricultural cooperatives affect the adoption of IPM technology by smallholder farmers. Moreover, IPM adoption may influence agricultural performance of farm households. For instance, adoption of IPM technology may increase farm profitability since it saves production costs. Understanding the issue is of great importance, since the effectiveness of agricultural policies that promote IPM adoption might be improved by taking into account the economic performance of farm households. However, much less is known about the farm-level economic performance associated with IPM adoption. This study attempts to fill the research gap and contribute to the literature in threefold.

First, we employ an endogenous switching probit (ESP) model to address the issue of selection bias in the process of choosing cooperative membership, and analyze the impact of cooperative membership on IPM adoption. The decision to join an agricultural cooperative is not a random event and depends on a number of observable factors (e.g. age, education and farm size) and unobservable factors (e.g. farmers’ innate abilities and motivations to enhance food safety and improve environmental performance). Although previous studies have employed propensity score matching (PSM) method to analyze the causal effect of cooperative membership on agricultural technology adoption, the approach addresses the self-selection issue by accounting for only observable factors, but fails to capture the factors that influence farmers’ decisions to adopt IPM.

Second, we employ a treatment effects model to analyze the impact of IPM adoption on crop yields, net returns and agricultural income. The treatment effects model adjusts for heterogeneity of IPM adoption by taking into consideration covariates affecting selection bias. Fernandez-Cornejo (1996) employed a standard Heckman two-step model to analyze the impact of IPM adoption on pesticide use, tomato yields and farm profits. However, the standard Heckman model emphasizes modeling structures of selection bias rather than assuming mechanisms of randomization work to balance data between IPM adopters and non-adopters.

Third, we employ a control function approach to address the potential endogeneity of off-farm work variable in cooperative membership choice equation and IPM adoption equation. Although income from off-farm activities help relax farmers’ financial constraints and enhance investment in yield-enhancing and quality-improving technologies (Matshe and Young, 2004), allocating more labor to cooperative activities would result in less time being allocated to off-farm work. Similarly, IPM adoption may be influenced by off-farm work participation, since additional income acquired from off-farm work activities also enables farmers to purchase more agricultural inputs used in the IPM (e.g. yellow sticky mobile, fixed traps, insect-trap light and trap band). However, allocating household labor and capital to off-farm work may constrain sustainable IPM technology management, resulting in the lost-labor effect in agricultural production as emphasized in the new economics of labor migration literature (Shi et al., 2011; Taylor et al., 2003). The joint relationship between off-farm work participation, cooperative membership choice and IPM adoption suggests potential endogeneity of off-farm work, which should be addressed for consistent estimation.

The study utilizes a cross-sectional survey data of 481 households in three major apple producing provinces (Gansu, Shaanxi and Shandong) in China. The apple sector in China is an interesting example, because being the largest apple producer in the world, China produces almost half of the world’s total apple output. However, only 3 percent of apples produced in the country are exported due to pesticide residual issues (FAOSTAT). The total pesticide use has increased from 0.73m tons in 1990 to 1.81m tons in 2012, while the total pesticide expenditure rose more than 11-fold between this period (CRSY, 2013). Particularly, fruit and vegetable production is intensive in pesticides, and apple is no exception. The rising use of chemical pesticides has increased farmers’ production costs and caused serious food safety, health and environmental problems. IPM technology is therefore being promoted intensively among apple producers to help reduce these adverse environmental impacts of conventional agriculture.

This paper proceeds as follows. Section 2 presents the analytical framework and estimation technique. We present the data and descriptive statistics in Section 3. The empirical results and discussion are presented in Section 4, and the final section concludes.

2. Analytical framework and estimation technique

2.1 Cooperative membership choice decision

Following Ito et al. (2012) and Abebaw and Haile (2013), the decision to join an agricultural cooperative is modeled in a random utility framework. Let C i * denote the difference between the utility derived from choosing cooperative membership ( C 1 i * ) and that derived from choosing the non-membership ( C 0 i * ), such that a household i will choose to join an agricultural cooperative, if C i * = C 1 i * C 0 i * > 0 . However, this difference is unobservable, but can be expressed by a latent variable model as follows:

(1) C i * = ψ Z i + ε i > 0 with C i = 1 if C i * > 0 ,
where Ci=1 if a farmer joined an agricultural cooperative and Ci=0 otherwise; Zi refers to a vector of variables (e.g. age, sex, farm size, household size and off-farm work participation) that may affect cooperative membership choice; ψ is a vector of parameters to be estimated; and εi is an error term, which is assumed to be normally distributed with zero mean.

Agricultural cooperatives may disseminate IPM knowledge among members, resulting in different adoption rates of the technology between cooperative members and non-members. For the analytical setting, let Y i * represent the net benefits acquired from adopting IPM technology, we observe Yi, if the underlying latent variable Y i * exceeds a certain threshold. Given our interest of exploring the impact of cooperative membership on IPM adoption, while controlling for other factors that may influence farmers’ decisions to adopt, we express farmers’ IPM adoption decisions as a latent variable function:

(2) Y i * = α X i + η C i + ϑ i with Y i = 1 if Y i * > 0 ,
where Y i * is a latent variable that represents the propensity to adopt IPM technology for household i, which gives the value of 1, if the farmer adopts IPM technology and 0 otherwise; Xi is a vector of observable characteristics (e.g. age, education, household size and off-farm work participation) that are assumed to influence IPM adoption; Ci is an indicator representing the farmer’s binary choice of cooperative membership; α and η are parameters to be estimated; and ϑi is a random error term.

Considering that farmers themselves decide (self-selection) whether to join a cooperative, the coefficient η that captures the impact of cooperative membership on IPM adoption may be biased. Ito et al. (2012) and Verhofstadt and Maertens (2014) used PSM method to account for such selection bias. However, PSM addresses selection bias depending on observable factors. When there are unobservable factors (e.g. farmers’ innate abilities) that simultaneously influence farmers’ decisions to choose cooperative membership and their IPM adoption decisions, PSM approach may still result in biased estimates. In light of the above problem, the present study employs an ESP model to address sample selection issues (Lokshin and Sajaia, 2011).

2.2 Modeling the impact of cooperative membership on IPM adoption

ESP model consists of two stages. The first stage models farmers’ decisions to choose cooperative membership, measured by Equation (1) and a probit model. In the second stage, a probit model is used to examine the relationship between IPM adoption variable and a set of explanatory variables conditional on the choice of cooperative membership. The two outcome equations, conditional on the choice of cooperative membership, can be expressed as follows:

(3a) Y 1 i * = β 1 X 1 i + ξ 1 i with Y 1 i = { 1 if Y 1 i * > 0 0 if Y 1 i * 0 if C i = 1 ,
(3b) Y 0 i * = β 0 X 0 i + ξ 0 i with Y 0 i = { 1 if Y 0 i * > 0 0 if Y 0 i * 0 if C i = 0 ,
where Y 1 i * and Y 0 i * are two latent IPM adoption variables for cooperative members and non-members, respectively; Y1i and Y0i are observed adoption choices, which take the value of 1 if cooperative members and non-members adopt the IPM technology, and 0 otherwise; Xi is a vector of observable variables (e.g. age, education, household size and off-farm work) that affect the decision to adopt IPM; β1 and β0 are parameters to be estimated; ξ1i and ξ0i are two error terms that represent unobservable factors related to IPM adoption for members and non-members, respectively. The full information maximum likelihood approach estimates the selection Equation (1) and outcome Equations (3a) and (3b) simultaneously (Ayuya et al., 2015; Lokshin and Sajaia, 2011).

For ESP model identification, a variable representing a farmer’s perception whether contemporary agricultural cooperative is more effective than people’s commune system (PCS) is used as an instrumental variable. PCS was a collective farming regime that was practiced during the 1950s–1970s, which finally resulted in production stagnation. Ito et al. (2012) found that farmers’ cooperative membership choice decision was significantly influenced by the image of the PCS. To test the validity of the PCS perception variable as an instrument, we run simple probit models for the cooperative membership choice equation and the IPM adoption equation with the inclusion of the instrumental variable as a regressor. The results, which are not presented for the sake of brevity, but are available on request, show that the coefficients of the PCS perception variable is positive and significant in the cooperative membership choice specification, but statistically insignificant in IPM adoption specification. Furthermore, Pearson correlation analysis also reveals that the PCS perception variable is significantly correlated with the cooperative membership variable, but uncorrelated with the IPM adoption variable. The findings confirm the validity of the PCS perception variable as an instrument.

In addition to exploring important factors that influence farmers’ decisions to choose cooperative membership and the determinants of IPM adoption separately for cooperative members and non-members, we are interested in the treatment effects of cooperative membership on IPM adoption. In particular, the average treatment effects on the treated (ATT) and average treatment effects on the untreated (ATU) are of interest. Following Lokshin and Sajaia (2011), the ATT and ATU can be calculated as follows:

(4a) ATT = 1 N 1 i = 1 N 1 [ Pr ( Y 1 = 1 | C = 1 , X = x ) Pr ( Y 0 = 1 | C = 1 , X = x ) ] ,
(4b) ATU = 1 N 0 i = 1 N 0 [ Pr ( Y 1 = 1 | C = 0 , X = x ) Pr ( Y 0 = 1 | C = 0 , X = x ) ] ,
where N1 and N0 represent the sample numbers of cooperative members and non-members, respectively; Pr(Y1=1|C=1, X=x) and Pr(Y0=1|C=0, X=x) are predicted probabilities of IPM adoption for cooperative members and non-members in an observed context, while Pr(Y0=1|C=1, X=x) and Pr(Y1=1|C=0, X=x) are predicted IPM technology probabilities for those two groups of farmers in a counterfactual context, respectively.

As discussed earlier, the joint relationship between off-farm work participation, cooperative membership choice and IPM adoption suggests potential endogeneity of off-farm work in cooperative membership choice Equation (1) and IPM adoption Equations (3a) and (3b). Ignoring the endogeneity associated with off-farm work participation would produce biased estimates. In this study, a control function approach proposed by Wooldridge (2015) is used to address the problem of endogenous off-farm work variable in estimating the cooperative membership choice specification and the IPM adoption specifications.

The control function approach includes two stages. In the first stage, the off-farm work variable is specified as a function of all other explanatory variables with the inclusion of one instrumental variable. A variable representing a farmer’s perception whether it is difficult to find an off-farm work is used as an identifying instrument. We employed the same strategy as we used for testing the validity of the PCS perception variable to check the validity of the instrumental variable for off-farm work. In the second stage regression, the residual predicted from the first stage is included as an additional regressor in cooperative membership choice Equation (1) and IPM adoption Equations (3a) and (3b).

2.3 Modeling the impact of IPM adoption on farm economic performance

The ESP model estimation can provide significant understanding on the determinants of IPM adoption and the extent to which agricultural cooperatives may affect IPM adoption. In addition to benefiting environment, human health and food quality, IPM adoption is expected to maintain or enhance farm output. Thus, we also examine the impact of IPM adoption on economic performance of farm households. The relationship for examining the impact of IPM adoption on farm performance assumes a linear specification for farm performance indicator (e.g. apple yields, net returns or agricultural income) as a function of a vector of explanatory variables (Xi) along with a dummy variable for IPM adoption (Yi). The regression equation for farm performance (Hi) can be specified as:

(5) H i = ω Y i + ν X i + δ i ,
where Hi represents farm performance indicators such as crop yields, net returns or agricultural income; Yi is a 0 or 1 dummy variable for IPM adoption; Xi summarizes observed individual and household characteristics (e.g. age, education, farm size, household size and asset ownership) that may influence the farm performance indicators; ω and ν are parameters to be estimated; and δi is an error term.

As indicated earlier, farmers choose to adopt IPM technology themselves. Farmers who choose to adopt IPM technology are therefore likely to have characteristics that could allow them to be more successful in using the technology than the average farmers. Therefore, due to self-selection bias, it would be incorrect to employ OLS regression to directly estimate the impact of IPM adoption on economic performance indicators, as in Equation (5). In this study, we employ a treatment effects model to analyze the impact of IPM adoption on farm performance indicators.

By using the treatment effects model, the selection mechanism for IPM adoption by the probit model can be explicitly specified as follows:

(6) Y i * = ζ X i + κ T i + ϖ i with Y i = 1 if Y i * > 0 ,
where Y i * is a latent variable and Yi is its proxy variable in an observed context, Xi is a vector of explanatory variables (e.g. age, education, household size and off-farm work participation), as defined in Equation (2); ζ and κ are parameters to be estimated; and ϖi is a random error term; Ti is a vector of instrumental variables that are expected to influence farmers’ decisions to adopt IPM technology, but do not affect apple yields, net returns and agricultural income. In this study, the variables information availability and benefit availability are used as instruments in the analysis.

3. Data and descriptive statistics

3.1 Data

The data used in this paper come from a farm household survey of apple farmers conducted between September and December 2013 in China. A multi-stage sampling procedure was used to select sample households for the empirical analysis. In the first stage, Gansu, Shaanxi and Shandong provinces were selected, because they are the top three apple producing regions with respect to the orchard areas cultivated in the country. In particular, the apple orchards in Gansu, Shaanxi and Shandong provinces are 283.9, 645.2 and 279.6 thousand hectares, respectively, contributing 54.17 percent of the country’s total apple orchards in 2012 (CRSY, 2013). In the second stage, Jingning county in Gansu, Luochuan county in Shaanxi, and Qixia and Laiyang cities in Shandong, where apple production is intensive at the provincial level, were chosen. Third, six agricultural cooperatives were randomly selected from those districts, using the information provided by the local agricultural bureaus in each purposively selected province. Fourth, three villages affiliated to each cooperative were randomly selected. In the last stage, in each village we randomly selected 25–30 households including both cooperative members and non-members. The procedure results in a total of 481 households, with 208 having cooperative membership and 273 having no membership.

We focus on agricultural cooperatives specialized in apple production and marketing in the present study, because they share similar attributes in helping members across different provinces. With respect to apple production, the cooperatives are responsible for enhancing adoption and diffusion of agricultural technologies, including IPM technology. The cooperatives are open to all individual households who are able to use their services and willing to accept the responsibilities of membership. The development of agricultural cooperatives and members’ legal rights are supported by the Farmers’ Professional Cooperatives Law in China. We gathered information from selected households through pre-tested questionnaire interviews. The questionnaire covered a range of topics including socioeconomic and farm-level factors, IPM practices, yields, gross income and production costs associated with apple production, off-farm work, income from other farm activities, farmers’ environmental and health perceptions associated with the continuous use of chemical pesticides, as well as asset ownership.

In this study, the first objective aims to analyze the impact of cooperative membership on IPM adoption. However, as noted by Fernandez-Cornejo (1996), the development of IPM programs is so different across pest class, crops and regions that it is difficult to provide a general operational definition of IPM. Our operational definition of IPM adoption follows the studies by Fernandez-Cornejo (1996) and Dasgupta et al. (2007), who defined IPM adoption as a dichotomous decision. Specifically, a farmer is defined as an IPM adopter: if the farmer reports having used both scouting for pests and economic thresholds in making pest treatment decisions; if the farmer reports adjusting application rates, time and frequency of pesticide use; and if the farmer uses any of the following methods: yellow sticky mobile, fixed traps, insect-trap light, trap band, cardboard traps to target adult, purchasing beneficial insects that prey on insects damaging to the crop. IPM non-adopters refer to those farmers who only depend on pesticides for pest management[1].

The second objective of this study is to analyze the impact of IPM adoption on apple yields, net returns and agricultural income. In particular, apple yields refer to apple yields per mu (1 mu=1/15 hectare). Net returns measure the difference between gross income of apple yields and variable costs (including fertilizers, pesticides, bags, irrigation, hired labors and agricultural films) per mu. Agricultural income measures per capita agricultural income. The total agricultural income includes the income from apple production and the income from other farm activities such as raising livestock and growing other crops such as potatoes, pears, peaches, cherries, peanuts, corn and apricot.

3.2 Descriptive statistics

The definition and descriptive statistics for the variables used in the empirical analysis are presented in Table I. The independent variables were selected based on past studies on determinants of cooperative membership and IPM adoption (e.g. Abebaw and Haile, 2013; Carrión Yaguana et al., 2016; Dasgupta et al., 2007; Verhofstadt and Maertens, 2014). The survey showed that the households with cooperative membership represented 43 percent of the total sample. The average apple yields and net returns are 2,220 kg/mu and 7,540 yuan/mu, respectively. The mean value of agricultural income is 13,460 yuan per capita. Only 21 percent of households adopted IPM technology, showing a lower adoption rate. The average number of years of schooling of the household head is about 7.6 years. Around 15 percent of surveyed household heads participated in off-farm work, suggesting agricultural production is the primary profession for most of the farmers.

Table II presents differences in means in the characteristics of cooperative members and non-members. In particular, cooperative members are more educated than non-members. They have larger farm sizes, and are more likely to own assets such as farming vehicle. Compared with non-members, cooperative members are more likely to believe that food produced under organic food standard, green food standard or pollution-free food standard receives higher prices than conventional food[2]. The mean comparisons in Table II also show that cooperative members and non-members are also distinguishable in terms of environmental perception and health perception. In particular, members are more likely to be aware of negative health and environmental effects associated with continuous use of chemical pesticides than non-members. With regards to the variable that represents IPM adoption, we find that cooperative members are more likely to adopt IPM technology than nonmembers.

Table III presents the mean differences in characteristics between IPM adopters and non-adopters. It shows that IPM adopters are younger than non-adopters. IPM adopters are better educated, and have larger farm sizes, compared with non-adopters. The mean comparisons indicate that IPM adopters and non-adopters are significantly different with respect to off-farm work participation, price knowledge, environmental and health perception. Table III also shows that apple yields for IPM adopters are significantly lower than that for non-adopters. There is, however, no significant difference in net returns between IPM adopters and non-adopters. With regard to agricultural income, the descriptive analysis in Table III shows that agricultural income obtained by IPM adopters are significantly higher than that received for non-adopters.

Overall, the descriptive statistics presented in Tables II and III show that cooperative members and non-members, IPM adopters and non-adopters are systematically different in observed household and farm-level characteristics. However, given that farmers choose to join agricultural cooperatives themselves, the differences in IPM adoption between cooperative members and non-members presented in Table II are inconclusive with respect to understanding the impact of cooperative membership on IPM adoption. Moreover, because farmers self-select into IPM programs, the differences in apple yields, net returns and agricultural income between IPM adopters and non-adopters presented in Table III are not sufficient to help understand the impact of IPM adoption on farm outcomes of interest. Thus, rigorous impact assessment methods such as ESP model and treatment effects model are, respectively, employed to estimate the impact of cooperative membership on IPM adoption, as well as the effect of IPM adoption on apple yields, net returns and agricultural income.

4. Results and discussion

4.1 Determinants of cooperative membership

The ESP estimation results for the determinants of cooperative membership, which are estimated from the first stage of ESP model, are presented in the second column of Table IV. The results show that the probability of being a member in a cooperative significantly increases with household education, a finding that is consistent with the findings in previous studies (Abate et al., 2014; Ito et al., 2012; Ma and Abdulai, 2016; Mojo et al., 2017). This is consistent with the expectation that the probability of choosing cooperative membership increases with the level of education of the household head due to greater awareness of the benefits associated with agricultural cooperatives. Our finding, consistent with Mojo et al. (2017), shows that farmers with larger household size and farm size are more likely to join cooperatives. Ownership of farming vehicle and access to warehouse appear to be important determinants of cooperative membership. The coefficients of the variables representing environmental perception and health perception are positive and significantly different from zero, suggesting that farmers who consider continuous use of chemical pesticides as a threat to environmental protection and human health are more likely to have cooperative membership. Previous studies have shown that cooperative organizations help improve members’ food safety performance through technical assistance (Moustier et al., 2010; Naziri et al., 2014), while the technologies suggested by cooperatives may benefit the environment and health. Relative to farmers in Shandong, farmers in Shaanxi are less likely to join cooperatives.

Note that the primary objective of the selection equation in ESP model is not to perfectly explain the determinants of cooperative membership, but to account for unobserved heterogeneity that could bias the treatment effects of cooperative membership on IPM adoption. For this purpose, the selection equation needs to include at least one valid instrument, which should be excluded in the outcome equations. As noted by Lokshin and Sajaia (2011), missing instrumental variable in the selection equation may make the ESP model be identified by non-linearities. The estimated results show that the PCS perception variable that served as the instrumental variable is positive and significantly different from zero, suggesting that farmers who perceive that contemporary agricultural cooperative is more effective than PCM are more likely to choose cooperative membership.

4.2 Determinants of IPM adoption

The ESP estimation results for the IPM adoption equations of cooperative members and non-members are shown in the third and fourth columns of Table IV. The coefficient estimates for the cooperative member and non-member regimes differ notably with respect to some of the variables. The estimates of the residuals of the off-farm work variable, derived from the first-stage regression of off-farm work in control function approach, are not significantly different from zero in the IPM adoption specifications for cooperative members and non-members, suggesting that there is no simultaneity bias and that the coefficients are consistently estimated (Wooldridge, 2015).

The results in Table IV show that age is an important factor in explaining lower probability of adopting IPM technology among cooperative members. This is probably due to the fact that older farmers are more used to traditional methods for pest management rather than learning the knowledge-intensive IPM technology. The negative and significant coefficient of sex variable for nonmembers suggests that male non-members are less likely to adopt IPM technology. The estimated coefficient of education variable shows a positive and statistically significant effect on the probabilities of IPM adoption for both cooperative members and non-members, suggesting that well-educated farmers are more likely to adopt IPM technology. Good knowledge makes farmers better able to understand the importance and benefits associated with IPM technology. The finding on education is consistent with other empirical studies on the effect of education on the adoption of IPM strategies (e.g. Dasgupta et al., 2007; Fernandez-Cornejo, 1998). The estimates reveal that farm size has a positive and significant impact on the probability of adopting IPM technology for cooperative members, but a negative and insignificant effect on IPM adoption for non-members. Larger farm size may help members obtain higher benefits from IPM adoption, contributing to a higher likelihood of IPM adoption. However, IPM adoption may involve risks such as productivity loss for non-members in the absence of technical assistance, resulting in decreasing adoption likelihood with increasing farm size for this group of farmers. The variable representing household size has a positive and significant effect on IPM adoption for members, suggesting that larger households with potentially more labor supply are more likely to adopt labor-intensive IPM technology under the cooperatives’ guidance.

The coefficients of the variable representing off-farm work are positive for both cooperative members and non-members, but only statistically significant for non-members. The positive and significant impact is consistent with the income effect of off-farm work participation, since off-farm earnings help farmers overcome credit and insurance market constraints by providing liquidity for the purchase of equipment such as fixed traps, insect-trap light and yellow stick mobile for IPM technology. However, the finding contradicts with the result reported by Shi et al. (2011), who found that income effect of off-farm work participation cannot compensate for the lost-labor effect, contributing to a negative relationship between off-farm work participation and the levels of chemical input use in rice production in Jiangxi province of China. Ownership of assets such as farming vehicle and the availability of apple refrigerated warehouse appears to increase the probability of IPM adoption for cooperative members. The positive and significant coefficient of price knowledge variable for cooperative members suggests that farmers who perceive that food produced under food safety standards obtains a higher price than conventional food are more likely to adopt IPM technology. To the extent that IPM adoption improves food safety due to the reduction of chemical pesticide use, the high-quality product is expected to obtain favorable price (Moustier et al., 2010; Naziri et al., 2014).

The coefficients of the environmental perception variable are positive and statistically significant, suggesting that farmers who are aware of pesticide pollution on the environment are more likely to adopt IPM technology. Cooperative members with health perception associated with chemical pesticide use tend to have a higher probability of adopting IPM technology. The finding supports the study by Dasgupta et al. (2007) who found that farmers who attribute their poor health to pesticide use may be more likely to adopt IPM, since IPM adoption may well improve health. The results in Table IV also reveal that location-fixed effects may be significant in explaining differences in IPM adoption. In particular, non-members located in Shaanxi are more likely to adopt IPM technology. The significance of location variable reflects variation in cropping systems, temperature, humidity and regional pest populations in determining farmers’ decisions to adopt IPM technology.

In the lower part of Table IV, we present the estimates of correlation coefficients (ρ0 and ρ1) of covariance terms between the error term in Equation (1) and the error terms in the outcome Equations (3a) and (3b). The significance of ρ0 confirms the presence of selection bias arising from unobservable factors, suggesting that addressing the self-selection bias issue by accounting for both observable and unobservable factors is a prerequisite for obtaining consistent and unbiased treatment effect of cooperative membership on IPM adoption (Ayuya et al., 2015; Gregory and Coleman-Jensen, 2013; Lokshin and Sajaia, 2011). Moreover, the results also show that the Wald test of the joint significance of the correlation coefficient rejects the null hypothesis that there is no correlation between cooperative membership choice equation and IPM adoption equation, indicating that it is more efficient to use the ESP model than a simple probit model. Finally, the estimates of the residuals of the off-farm work variable, derived from the first-stage regression of off-farm work in control function approach, are not significantly different from zero, suggesting that there is no simultaneity bias and that the coefficients are consistently estimated (Wooldridge, 2015).

4.3 Estimating average treatment effects

We now use the estimated coefficients from the ESP model in combination with Equations (4a) and (4b) to calculate the average treatment effects (ATT and ATU) of cooperative membership on IPM adoption. The results are presented in Table V. Unlike the mean values predicted in Table II, these ATT and ATU estimates account for selection bias arising from both observable and unobservable factors. Specifically, the ATT estimate shows that the causal effect of cooperative membership was to significantly increase the probability of adopting IPM technology by 29.5 percent. The ATU estimate in Table V is also statistically significant, which suggests that farmers without cooperative membership would be 9.7 percent more likely to adopt IPM technology if they were involved in cooperative organizations.

To gain further understanding of the impact of cooperative membership on IPM adoption, we also present in Table V the ATT and ATU estimates based on surveyed regions. The results generally show that the causal effects of cooperative membership were to increase the probabilities of adopting IPM technology, and agricultural cooperative in Gansu plays the largest effect. In particular, cooperative members in Gansu are 50.7 percent more likely to adopt IPM technology. For cooperative members in Shaanxi and Shandong, the casual effects of cooperative membership were to increase the likelihoods of IPM adoption by 39.1 and 12.8 percent, respectively. Overall, the findings of positive relationship between cooperative membership and IPM adoption in Table V generally suggest that agricultural cooperative can be a transmission route in the efforts to spread IPM technology that may contribute to the improvements of food safety, health and environmental performance.

4.4 Impact of IPM adoption on farm economic performance

The empirical results of the impact of IPM adoption on farm economic performance indicators such as apple yields, net returns and agricultural income are presented in Table VI. As indicated previously, treatment effects model was employed to jointly estimate the IPM adoption equation and three farm performance equations, respectively.

The lower parts of Table VI show that the estimates of the correlation coefficients ρδϖ in Models 1–3 are significantly different from zero, suggesting the presence of selection bias due to unobservable factors. In particular, the negative correlation coefficients of ρδϖ indicate negative selection bias. This would suggest that farmers having lower than average apple yields, net returns and agricultural income have higher probabilities of adopting IPM technology. In other words, farmers who expect that IPM adoption may improve the efficiency of pest management and save production costs are more likely to be IPM adopters, since the improvement of pest management efficiency may contribute to increased yields and the reduction of investment costs is closely associated with higher net returns and agricultural income. Thus, failing to account for such negative selection bias would lead to underestimated effects of IPM adoption on apple yields, net returns and agricultural income. The results of the Wald tests for ρδϖ = 0 in Models 1–3 are significantly different from zero, suggesting that the null hypothesis that the IPM adoption variable is exogenous in three farm performance equations can be rejected.

The results from the first-stage estimates of the treatment effects model, which show the determinants of farmers’ decisions to adopt IPM technology, are presented in the second, fourth and sixth columns in Table VI. The potential endogeneity of off-farm work variable in IPM adoption equation is also addressed using the control function approach proposed by Wooldridge (2015). The statistically insignificant coefficients of the off-farm work residual variables in Models 1–3 in Table VI also confirm the exogeneity of off-farm participation in IPM adoption equations (Wooldridge, 2015). Given that the primary objective of IPM adoption equation estimation is to account for unobserved heterogeneities that may bias the impact of IPM adoption on outcomes of interest, detailed interpretation of the results are not provided here. Next, we mainly discuss the impact of IPM adoption and other control variables on farm economic performance indicators such as apple yields, net returns and agricultural income, based on the results presented in Columns 3, 5 and 7 in Table VI.

Focusing on the variable of primary interest, IPM adoption, controlling for selection bias arising from observable and unobservable factors has positive and statistically significant impacts on apple yields, net returns and agricultural income. Such effects could not be observed when only comparing descriptive statistics between IPM adopters and non-adopters, due to the mentioned negative selection bias. The findings of positive impact of IPM adoption on apple yields and net returns are consistent with the finding by Fernandez-Cornejo (1998), who found that IPM adoption for diseases significantly increases yields and profits for grape growers in the USA. Dasgupta et al. (2007) also found that IPM rice farming is more profitable than conventional rice farming, while there is no significant productivity difference between IPM and conventional rice farming.

The other coefficient estimates in Table VI show that apple yields, net returns and agricultural income are also affected by several other factors. The coefficients of age variable in Models 2 and 3 are negative and statistically significant, suggesting that older farmers obtain significantly lower net returns and agricultural income. The positive and significant coefficients of sex variable in Table VI suggest that male household heads obtain higher apple yields, net returns and agricultural income, compared with their female counterparts. Farm size appears to have differential impacts on apple yields, net returns and agricultural income. The negative and significant coefficients of farm size in Models 1 and 2 suggest that larger farms obtain significantly lower apple yields and net returns than smaller farms. The finding is consistent with the results reported by Chen et al. (2011), who found a negative relationship between farm size and productivity in China. Farm size appears to have a positive and statistically significant impact on agricultural income.

The coefficient representing the effect of household size is positive and statistically significant in the case of apple yields but negative and significant for agricultural income. To the extent that large household size implies more labor endowment, which can enhance adoption of yield-enhancing technologies and finally contributes to higher apple yields. However, large household size reduces the per capita agricultural income. Participation in off-farm work tends to have a significant and negative effect on apple yields, net returns and agricultural income. The finding is consistent with the finding of lost-labor effect of off-farm work on farm economic performance (Taylor et al., 2003). Although income from off-farm work enables farmers to purchase production inputs, allocating more time to off-farm work results in less time being allocated to farm operations, leading to reduced efficiency and profitability of agricultural production. The coefficients of price knowledge are positive and significantly differently from zero in Table VI, suggesting that farmers who perceive that food produced under food safety standards obtains a higher price than conventional food obtain high apple yields, net returns and agricultural income.

Variables indicating environmental perception and health perception were found to influence outcomes, although at varying levels. The negative and significant coefficient of environmental perception variable in Column 3 suggests that farmers who consider continuous use of chemical pesticides as a threat to environmental performance obtain low apple yields. Environmental concern associated with chemical pesticides use may motivate farmers to use less agrochemicals, resulting in lower yields. Health perception variable in Column 7 has a positive and significant coefficient, suggesting that health concerns associated with pesticide use is correlated with high agricultural income. The results in Table VI also reveal that location-fixed effects are important in determining farm economic performance. In particular, compared with farmers located in Shandong (reference region), farmers located in Shaanxi tend to have lower apple yields, net returns and agricultural income. Moreover, farmers in Gansu also obtain significantly lower apple yields relative to their counterparts in Shandong. The significant influence of location variables indicates that accounting for unobserved agro-climatic, institutional and socioeconomic heterogeneities among the sample districts is a prerequisite for consistent estimates of farm economic performance.

Finally, we also estimated the impact of IPM adoption on chemical pesticide adoption, with the aim of testing whether IPM adoption enables farmers to minimize chemical pesticide use in relation to pest management. Given the variable for chemical pesticide adoption is dichotomous, a recursive bivariate probit model is estimated (Thuo et al., 2014). The results, which are presented in Table AI, show that IPM adoption exerts a negative and statistically significant impact on the probability of adopting chemical pesticide.

5. Conclusions and policy implications

Although IPM technology is promoted as a preferred approach for both sustainable intensification of crop production and pesticide risk reduction, its adoption rate remains low in China. There is still a need to facilitate the adoption of the technology. While recent studies have shown that agricultural cooperative is an efficient institutional innovation that enhances farmers’ adoption of agricultural technologies, there is hardly any work that has looked at the question whether agricultural cooperatives can promote IPM adoption, and how adoption impacts on agricultural performance. In this paper, we have addressed the research gap by analyzing the impact of cooperative membership on IPM adoption. We employed an ESP model that accounts for sample selection bias and structural differences between cooperative members and non-members to analyze the adoption behaviors of 481 apple farmers in China. To further understand how IPM adoption influences the economic performance of farm households, we employed a treatment effects model to analyze the impact of IPM adoption on apple yields, net returns and agricultural income.

The empirical results revealed the presence of selection bias arising from unobserved heterogeneities. After controlling for this bias, the ATT estimate showed that the causal effect of cooperative membership was to increase the probability of IPM adoption by 30 percent. On the other hand, the positive and significant ATU suggested that farmers without cooperative membership would be 10 percent more likely to adopt IPM technology if they joined cooperatives. Overall, our results suggest that agricultural cooperative could be an important transmission route in the government’s efforts of spreading IPM technology.

With respect to the relationship between IPM adoption and farm performance indicators, the simple mean value comparisons revealed apple yields of IPM adopters were significantly lower than that of non-adopters, and there was no significant difference in net returns between these two groups of farmers. However, econometric estimation with a treatment effects model revealed negative selection bias, implying that farmers with lower than average apple yields, net returns and agricultural income are more likely to adopt IPM technology. Controlling for this bias resulted in positive and significant impacts of IPM adoption on apple yields, net returns and agricultural income. The further estimation showed that IPM adoption significantly reduces the probability of adopting chemical pesticide.

Generally, the empirical results presented in this paper support the notion that membership in agricultural cooperatives can play a positive role by serving as a catalyst for spreading IPM technology, and increased IPM adoption tends to improve the economic performance of farm households. The finding that price knowledge tends to influence farmers’ decisions to join cooperatives and adopt IPM technology suggests that the enhancement of farmers’ price knowledge about food produced under organic food standard, green food standard and pollution-free food standard and the establishment of food safety markets would step up the promotion of farmers’ decisions to join cooperatives and adopt IPM technology. In particular, the Chinese government should continue to encourage the adoption of food safety and quality standards by agricultural cooperatives and enhance market mechanism of high price for high-quality products. The positive and significant impacts of environmental perception on IPM adoption suggest that promoting effective measures to improve farmers’ understanding of negative environmental effects associated with continuous use of chemical pesticides would help increase farmers’ adoption of IPM technology. This could be achieved through cooperatives’ collective activities. Finally, the positive impact of IPM adoption on apple yields, net returns and agricultural income underscores the importance of efforts by policy makers to promote IPM adoption through agricultural cooperatives.

Definition of variables and descriptive statistics

Variable Definition Mean (SD)
Dependent variables
Membership 1 if farmer had agricultural cooperative membership, 0 otherwise 0.43 (0.50)
IPM adoption 1 if farmer adopted integrated pest management (IPM) technology, 0 otherwise 0.21 (0.41)
Apple yields Apple output (1,000 kg/mu)a 2.22 (8.20)
Net returns Apple gross revenue minus variable investments costs (1,000 yuan/mu)b 7.54 (3.91)
Agricultural income Per capita agricultural income (1,000 yuan/capita) 13.46 (7.50)
Pesticide adoption 1 if farmer adopted chemical pesticide, 0 otherwise 0.92 (0.27)
Independent variables
Age Age of the household head (years) 48.63 (10.25)
Sex 1 if farmer is male, 0 otherwise 0.86 (0.35)
Education Formal education of farmer (years) 7.60 (2.87)
Farm size Total size of fruiting apple orchards (mu) 5.07 (3.24)
Household size Number of people residing in household 4.33 (1.44)
Off-farm work 1 if farmer participates in non-farm work, 0 otherwise 0.15 (0.36)
Asset ownership 1 If farmer owns farming vehicle, 0 otherwise 0.92 (0.28)
Warehouse 1 if farmer reports there is apple refrigerated warehouse in local areas, 0 otherwise 0.54 (0.50)
Price knowledge 1 if farmer perceives that food produced under food safety standards (i.e. organic food standard, green food standard or pollution-free food standard) may be sold at a higher price than conventional food, 0 otherwise 0.43 (0.50)
Environmental perception 1 if farmer considers continuous use of chemical pesticides as a threat to environmental performance, 0 otherwise 0.47 (0.50)
Health perception 1 if farmer considers continuous use of chemical pesticides as a threat to human health, 0 otherwise 0.55 (0.47)
Gansu 1 if farmer is located in Gansu, 0 otherwise 0.17 (0.37)
Shaanxi 1 if farmer is located in Shaanxi, 0 otherwise 0.40 (0.49)
Shandong 1 if farmer is located in Shandong, 0 otherwise 0.43 (0.50)
PCS perception 1 if farmer perceives that contemporary agricultural cooperative is more effective than people’s commune system (PCS), 0 otherwise 0.42 (0.49)
Information availability The availability of IPM information determines my decision to adopt IPM technology (1=strongly disagree; 2=general; 3=agree; 4=strongly agree) 2.79 (0.96)
Benefit availability My tuition tells me that I can benefit from the adoption of IPM technology (1= strongly disagree; 2= general; 3= agree; 4= strongly agree) 1.95 (0.86)

Notes: a1 mu=1/15 hectare; b$1=6.14 yuan

Mean differences in characteristics between cooperative members and non-members

Variables Members (n=208) Non-members (n=273) Diff.
Age 48.45 (0.66) 48.78 (0.66) −0.326
Sex 0.89 (0.02) 0.84 (0.02) 0.051
Education 8.05 (0.17) 7.27 (0.19) 0.781***
Farm size 5.51 (0.24) 4.73 (0.18) 0.778***
Household size 4.57 (0.10) 4.14 (0.08) 0.433***
Off-farm work 0.14 (0.02) 0.16 (0.02) −0.013
Asset ownership 0.96 (0.01) 0.88 (0.02) 0.079***
Warehouse 0.57 (0.03) 0.52 (0.03) 0.048
Price knowledge 0.50 (0.03) 0.37 (0.03) 0.130***
Environmental perception 0.55 (0.03) 0.41 (0.03) 0.141***
Health perception 0.64 (0.03) 0.48 (0.03) 0.160***
PCS perception 0.60 (0.03) 0.29 (0.03) 0.312***
IPM adoption 0.32 (0.03) 0.13 (0.02) 0.190***

Notes: Standard errors in parentheses. ***p<0.01

Mean differences in characteristics between IPM adopters and non-adopters

Variables IPM adopters (n=108) Non-adopters (n=378) Diff.
Age 44.10 (0.88) 49.87 (0.53) −5.771***
Sex 0.87 (0.03) 0.86 (0.02) 0.017
Education 8.77 (0.27) 7.29 (0.15) 1.479***
Farm size 7.00 (0.34) 4.54 (0.15) 2.460***
Household size 5.05 (0.13) 4.13 (0.07) 0.919***
Off-farm work 0.29 (0.04) 0.11 (0.02) 0.178***
Asset ownership 0.95 (0.02) 0.91 (0.01) 0.044
Warehouse 0.71 (0.04) 0.35 (0.02) 0.360***
Price knowledge 0.70 (0.05) 0.50 (0.03) 0.196***
Environmental perception 0.80 (0.04) 0.38 (0.02) 0.418***
Health perception 0.84 (0.04) 0.47 (0.03) 0.376***
Information availability 3.06 (0.08) 2.71 (0.05) 0.344***
Benefit availability 2.35 (0.09) 1.84 (0.04) 0.514***
Apple yields 1.93 (0.08) 2.24 (0.04) −0.306***
Net returns 7.15 (0.39) 7.65 (0.20) −0.494
Agricultural income 15.38 (0.86) 12.94 (0.36) 2.441***
Pesticide adoption 0.85 (0.03) 0.94 (0.01) −0.085***

Notes: Standard errors in parentheses. ***p<0.01

Determinants of cooperative membership and determinants of IPM adoption: ESP model estimation

IPM adoption
Variable Selection Members Non-members
Age −0.000 (0.008) −0.042 (0.016)** 0.013 (0.015)
Sex 0.165 (0.191) 0.623 (0.434) −0.919 (0.299)***
Education 0.049 (0.027)* 0.197 (0.074)*** 0.139 (0.056)**
Farm size 0.065 (0.029)** 0.140 (0.052)*** −0.065 (0.041)
Household size 0.182 (0.054)*** 0.290 (0.128)** −0.005 (0.107)
Off-farm work −0.441 (0.414) 2.877 (2.054) 1.656 (0.665)**
Asset ownership 0.670 (0.258)*** 1.283 (0.695)* 0.013 (0.320)
Warehouse 0.123 (0.137) 0.682 (0.330)** 0.299 (0.245)
Price knowledge 0.316 (0.144)** 0.885 (0.302)*** 0.549 (0.271)**
Environmental perception 0.239 (0.141)* 0.918 (0.476)* 0.655 (0.250)***
Health perception 0.237 (0.136)* 1.933 (0.669)*** 0.241 (0.212)
Shaanxi −1.043 (0.272)*** 0.490 (0.786) 1.521 (0.522)***
Gansu −0.331 (0.280) 0.916 (0.713) 0.455 (0.537)
Residual (off-farm work) 0.105 (0.120) 0.460 (0.687) −0.081 (0.196)
PCS perception 0.738 (0.130)***
Constant −2.672*** −8.025 (2.817)*** −4.283 (1.138)***
ρ1 (0.662) −0.167 (0.533)
ρ0 −1.521 (0.612)**
Log pseudo likelihood −382.089
Wald test of indep. eqns. (ρ1=ρ0) χ2 (2)=6.31, Prob>χ2=0.043
Observations 481 481 481

Notes: Robust standard errors in parentheses. The reference region is Shandong. *p<0.1; **p<0.05; ***p<0.01

Average treatment effects of cooperative membership on IPM adoption

Category ATT t-valuea ATU t-valueb
Full sample 0.295 (0.025)*** 11.93 0.097 (0.015)*** 6.58
Gansu 0.507 (0.053)*** 9.62 0.189 (0.050)*** 3.81
Shaanxi 0.391 (0.041)*** 9.472 0.089 (0.023)*** 3.94
Shandong 0.128 (0.030)*** 4.30 0.073 (0.019)*** 3.88

Notes: Standard errors in parentheses; a,bt-values are calculated based on the immediate form of the t-test command in Stata 13.1. ***p<0.01

Impact of IPM adoption on apple yields, net returns and agricultural income: treatment effects model estimation

Model 1 Model 2 Model 3
Variable IPM adoption Apple yields IPM adoption Net returns IPM adoption Agricultural income
IPM adoption 0.154 (0.091)* 0.272 (0.104)*** 0.222 (0.128)*
Age 0.006 (0.013) −0.003 (0.002) 0.006 (0.012) −0.008 (0.003)*** 0.012 (0.013) −0.007 (0.002)***
Sex −0.207 (0.277) 0.086 (0.045)* −0.214 (0.276) 0.130 (0.067)* −0.204 (0.268) 0.129 (0.067)*
Education 0.156 (0.048)*** −0.006 (0.006) 0.150 (0.047)*** −0.007 (0.010) 0.149 (0.048)*** −0.004 (0.009)
Farm size 0.069 (0.032)** −0.055 (0.006)*** 0.066 (0.031)** −0.055 (0.010)*** 0.040 (0.036) 0.127 (0.008)***
Household size 0.089 (0.094) 0.032 (0.011)*** 0.102 (0.084) 0.019 (0.019) 0.110 (0.088) −0.187 (0.017)***
Off-farm work 1.803 (0.624)*** −0.165 (0.051)*** 1.952 (0.579)*** −0.293 (0.073)*** 1.931 (0.576)*** −0.116 (0.066)*
Asset ownership 0.730 (0.319)** 0.055 (0.052) 0.721 (0.330)** 0.117 (0.082) 0.735 (0.326)** 0.052 (0.076)
Warehouse 0.729 (0.188)*** 0.048 (0.035) 0.748 (0.179)*** 0.062 (0.055) 0.682 (0.188)*** 0.072 (0.048)
Price knowledge 0.627 (0.180)*** 0.091 (0.031)*** 0.635 (0.179)*** 0.204 (0.049)*** 0.592 (0.177)*** 0.119 (0.045)***
Environmental perception 0.904 (0.219)*** −0.083 (0.035)** 0.903 (0.214)*** −0.083 (0.057) 0.957 (0.205)*** −0.064 (0.049)
Health perception 0.677 (0.211)*** −0.012 (0.032) 0.675 (0.205)*** −0.017 (0.053) 0.685 (0.206)*** 0.075 (0.045)*
Shaanxi 0.611 (0.419) −0.210 (0.054)*** 0.654 (0.403) −0.214 (0.094)** 0.818 (0.421)* −0.245 (0.070)***
Gansu 0.728 (0.398)* −0.290 (0.058)*** 0.770 (0.393)* 0.118 (0.094) 0.831 (0.385)** −0.100 (0.079)
Residual (off-farm work) 0.087 (0.195) 0.034 (0.173) 0.057 (0.186)
Information availability 0.284 (0.100)*** 0.291 (0.0968)*** 0.287 (0.095)***
Benefit availability 0.238 (0.127)* 0.253 (0.116)** 0.208 (0.111)*
Constant −7.574 (1.094)*** 7.942 (0.153)*** −7.690 (1.097)*** 9.175 (0.228)*** −7.795 (1.034)*** 9.721 (0.196)***
ρδϖ −0.333 (0.173)* −0.455 (0.107)*** −0.511 (0.174)***
Ln(σδϖ) −1.195 (0.034)*** −0.703 (0.037)*** −0.880 (0.047)
Wald test (ρδϖ=0) χ2 (1) =3.17, Prob>χ2=0.075 χ2 (1) =13.15, Prob>χ2=0.000 χ2 (1) =5.73, Prob>χ2=0.017
Observations 481 481 481

Notes: Robust standard errors in parentheses. The reference region is Shandong. The dependent variables in the second-stage estimation of treatment effects model include log form of apple yields measured in kg/mu in Model 1, log form of net returns from apple production measured in yuan/mu in Model 2, and log form of agricultural income measured in yuan/capita in Model 3. *p<0.1; **p<0.05; ***p<0.01

Impact of IPM adoption on chemical pesticide adoption

IPM adoption Chemical pesticide adoption
IPM adoption −1.011 (0.464)**
Age 0.003 (0.012) 0.001 (0.010)
Sex −0.102 (0.274) −0.126 (0.277)
Education 0.130 (0.048)*** −0.017 (0.045)
Farm size 0.069 (0.032)** −0.019 (0.031)
Household size 0.146 (0.085)* 0.069 (0.075)
Off-farm work 1.982 (0.615)*** 0.862 (0.390)**
Asset ownership 0.566 (0.347) −0.213 (0.352)
Warehouse 0.730 (0.193)*** 0.162 (0.210)
Price knowledge 0.673 (0.174)*** −0.027 (0.204)
Environmental perception 0.927 (0.216)*** 0.195 (0.206)
Health perception 0.831 (0.205)*** 0.036 (0.208)
Shaanxi 0.637 (0.421) −5.308 (0.361)***
Gansu 0.753 (0.394)* −4.940 (0.431)***
Residual (off-farm work) 0.032 (0.195)
Information availability 0.318 (0.104)***
Constant −7.203 (0.996)*** 6.465 (0.869)***
ρδϖ 0.615 (0.253)**
Wald test (ρδϖ = 0) χ2 (1)=3.095, Prob>χ2=0.079
Observations 481

Notes: Robust standard errors in parentheses. The reference region is Shandong. *p<0.1; **p<0.05; ***p<0.01

Notes

1.

Unlike the findings of lower adoption rates of pesticide reported by Abebaw and Haile (2013) on Ethiopian and Verhofstadt and Maertens (2014) on Rwanda, all apple farmers in our survey used different levels of pesticides (either chemical or biological pesticides, or both).

2.

In combination with the domestic agricultural practice and food safety situation, the Chinese Government proposed three food safety standards that include organic food standard, green food standard and pollution-free food standard (or known as safe food in some literature). In comparison with the unified international standard of organic food, the latter two safer food standards are unique in China. The requirements of these three safer food standards can be found in Yu et al. (2014).

Appendix

Table AI

References

Abate, G.T., Francesconi, G.N. and Getnet, K. (2014), “Impact of agricultural cooperatives on smallholders’ technical efficiency: evidence from Ethiopia”, Annals of Public and Cooperative Economics, Vol. 85 No. 2, pp. 257-286.

Abebaw, D. and Haile, M.G. (2013), “The impact of cooperatives on agricultural technology adoption: empirical evidence from Ethiopia”, Food Policy, Vol. 38, pp. 82-91.

Ayuya, O.I., Gido, E.O., Bett, H.K., Lagat, J.K., Kahi, A.K. and Bauer, S. (2015), “Effect of certified organic production systems on poverty among smallholder farmers: empirical evidence from Kenya”, World Development, Vol. 67, pp. 27-37.

Carrión Yaguana, V., Alwang, J., Norton, G. and Barrera, V. (2016), “Does IPM have staying power? Revisiting a potato-producing area years after formal training ended”, Journal of Agricultural Economics, Vol. 67 No. 2, pp. 308-323.

Chen, Z., Huffman, W.E. and Rozelle, S. (2011), “Inverse relationship between productivity and farm size: the case of China”, Contemporary Economic Policy, Vol. 29 No. 4, pp. 580-592.

Chhay, N., Seng, S., Tanaka, T., Yamauchi, A., Cedicol, E.C., Kawakita, K. and Chiba, S. (2017), “Rice productivity improvement in Cambodia through the application of technical recommendation in a Farmer Field School”, International Journal of Agricultural Sustainability, Vol. 15 No. 1, pp. 54-69.

CRSY (2013), China Rural Statistical Yearbook, China Statistics Press, Beijing.

Cuyno, L.C.M., Norton, G.W. and Rola, A. (2001), “Economic analysis of environmental benefits of integrated pest management: a Philippine case study”, Agricultural Economics, Vol. 25 Nos 2-3, pp. 227-233.

Dasgupta, S., Meisner, C. and Wheeler, D. (2007), “Is environmentally friendly agriculture less profitable for farmers? Evidence on integratedpest management in Bangladesh”, Applied Economic Perspectives and Policy, Vol. 29 No. 1, pp. 103-118.

Fernandez-Cornejo, J. (1996), “The microeconomic impact of IPM adoption: theory and application”, Annual Review of Resource Economics, Vol. 25 No. 2, pp. 149-160.

Fernandez-Cornejo, J. (1998), “Environmental and economic consequences of technology adoption: IPM in viticulture”, Agricultural Economics, Vol. 18 No. 2, pp. 145-155.

Fernandez-Cornejo, J., Jans, S. and Smith, M. (1998), “Issues in the economics of pesticide use in agriculture: a review of the empirical evidence”, Applied Economic Perspectives and Policy, Vol. 20 No. 2, pp. 462-488.

Gregory, C.A. and Coleman-Jensen, A. (2013), “Do high food prices increase food insecurity in the united states?”, Applied Economic Perspectives and Policy, Vol. 35 No. 4, pp. 679-707.

Ito, J., Bao, Z. and Su, Q. (2012), “Distributional effects of agricultural cooperatives in China: exclusion of smallholders and potential gains on participation”, Food Policy, Vol. 37 No. 6, pp. 700-709.

Kabir, M.H. and Rainis, R. (2014), “Adoption and intensity of integrated pest management (IPM) vegetable farming in Bangladesh: an approach to sustainable agricultural development”, Environment, Development and Sustainability, Vol. 17 No. 6, pp. 1413-1429.

Kelly, L. (2005), The Global Integrated Pest Management Facility, The World Bank, Washington, DC.

Lokshin, M. and Sajaia, Z. (2011), “Impact of interventions on discrete outcomes: maximum likelihood estimation of the binary choice models with binary endogenous regressors”, Stata Journal, Vol. 11 No. 3, pp. 368-385.

Ma, W. and Abdulai, A. (2016), “Does cooperative membership improve household welfare? Evidence from apple farmers in China”, Food Policy, Vol. 58, pp. 94-102.

Ma, W., Abdulai, A. and Goetz, R. (2018), “Agricultural cooperatives and investment in organic soil amendments and chemical fertilizer in China”, American Journal of Agricultural Economics, Vol. 100, February, pp. 502-520.

Matshe, I. and Young, T. (2004), “Off-farm labour allocation decisions in small-scale rural households in Zimbabwe”, Agricultural Economics, Vol. 30 No. 3, pp. 175-186.

Mojo, D., Fischer, C. and Degefa, T. (2017), “The determinants and economic impacts of coffee farmer cooperatives: recent evidence from rural Ethiopia”, Journal of Rural Studies, Vol. 50, pp. 84-94.

Moustier, P., Tam, P.T.G., Anh, D.T., Binh, V.T. and Loc, N.T.T. (2010), “The role of farmer organizations in supplying supermarkets with quality food in Vietnam”, Food Policy, Vol. 35 No. 1, pp. 69-78.

Naziri, D., Aubert, M., Codron, J.-M., Loc, N.T.T. and Moustier, P. (2014), “Estimating the impact of small-scale farmer collective action on food safety: the case of vegetables in Vietnam”, Journal of Development Studies, Vol. 50, January, pp. 715-730.

Sanglestsawai, S., Rejesus, R.M. and Yorobe, J.M. (2015), “Economic impacts of integrated pest management (IPM) Farmer Field Schools (FFS): evidence from onion farmers in the Philippines”, Agricultural Economics, Vol. 46 No. 2, pp. 149-162.

Shi, X., Heerink, N. and Qu, F. (2011), “Does off-farm employment contribute to agriculture-based environmental pollution? New insights from a village-level analysis in Jiangxi Province, China”, China Economic Review, Vol. 22 No. 4, pp. 524-533.

Taylor, J.E., Rozelle, S. and Brauw, A.D. (2003), “Migration and incomes in source communities: a new economics of migration perspective from China”, Economic Development and Cultural Change, Vol. 52 No. 1, pp. 75-101.

Thuo, M., Bell, A.A., Bravo-Ureta, B.E., Lachaud, M.A., Okello, D.K. et al. (2014), “Effects of social network factors on information acquisition and adoption of improved groundnut varieties: the case of Uganda and Kenya”, Agriculture and Human Values, Vol. 31 No. 3, pp. 339-353.

Van Den Berg, H. and Jiggins, J. (2007), “Investing in farmers – the impacts of Farmer Field Schools in relation to integrated pest management”, World Development, Vol. 35 No. 4, pp. 663-686.

Verhofstadt, E. and Maertens, M. (2014), “Smallholder cooperatives and agricultural performance in Rwanda: do organizational differences matter?”, Agricultural Economics, Vol. 45 No. S1, pp. 39-52.

Wilson, C. and Tisdell, C. (2001), “Why farmers continue to use pesticides despite environmental, health and sustainability costs”, Ecological Economics, Vol. 39 No. 3, pp. 449-462.

Wooldridge, J.M. (2015), “Control function methods in applied econometrics”, Journal of Human Resources, Vol. 50 No. 2, pp. 420-445.

Wossen, T., Abdoulaye, T., Alene, A., Haile, M.G., Feleke, S., Olanrewaju, A. and Manyong, V. (2017), “Impacts of extension access and cooperative membership on technology adoption and household welfare”, Journal of Rural Studies, Vol. 54, pp. 223-233.

Xie, W., Ali, T., Cui, Q. and Huang, J. (2017), “Economic impacts of commercializing insect-resistant GM maize in China”, China Agricultural Economic Review, Vol. 9 No. 3, pp. 340-354.

Yu, X., Gao, Z. and Zeng, Y. (2014), “Willingness to pay for the ‘Green Food’ in China”, Food Policy, Vol. 45, pp. 80-87.

Corresponding author

Wanglin Ma can be contacted at: Wanglin.Ma@lincoln.ac.nz